Image Credit: WR7/

Image Credit: WR7/

It’s never been more critical to harness advanced analytics for fleet maintenance.

Maintenance costs are skyrocketing, fueled by inflationary pressures, scarce replacement parts, and labor shortages. 2021 saw a 3.5% year-over-year increase in the cost of an average repair – and a nearly 12% jump in total vehicle maintenance cost, as wait times on new inventory keeps vehicles in service past their scheduled replacement dates. These trends have continued in 2022.

Industry commentators and vendors have heralded predictive maintenance as the answer to a looming industry crisis. But what does it take to finally go from idea to reality?


The promise of predictive maintenance

Predictive maintenance describes the practice of identifying hidden patterns in fleet vehicle data sources (telematics and repair, for example) to anticipate and avoid on-the-road failures. 

Imagine, for example, being able to perfectly predict failures for critical components at the individual vehicle level. You could move away from costly, one-size-fits-all preventative maintenance routines, instead servicing vehicles only as needed.

It’s an idea that resonates more than ever in a challenging economic climate. But fleets need adaptive maintenance to finally take it from theory to reality – and to unlock all the benefits of a data-driven approach to fleet maintenance.


The next frontier: adaptive maintenance

While predictive maintenance is likely familiar to folks, the idea of adaptive maintenance is still in its infancy.

Adaptive maintenance describes a set of data-driven strategies to lower maintenance cost, reduce downtime, and maximize long-term vehicle health and value. It differs from traditional predictive maintenance in two main ways:

  1. Analytical approach that meets fleets where they are. Every fleet can benefit from data-driven maintenance decisions. But not every organization has the same needs and capabilities around analytics. Adaptive maintenance ladders up in complexity based on where a fleet is – rather than imposing one-size-fits-all models on fleets with wildly divergent profiles.
  2. Broader focus and more business context. And adaptive maintenance moves past a reactive focus on preventing on-the-road failures. It’s about data-driven optimization for the long-term health and value of the vehicle – bearing in mind guardrails like regulatory compliance, warranty constraints, maintenance cost reduction, and parts and labor availability.

Adaptive maintenance solves three challenges with implementing predictive maintenance in practice:

  • Lack of precision: High-quality predictive models require data. Lots of it. Especially when forecasting relatively rare events (like on-the-road failures) – and even more so when doing so over hundreds or thousands of components and subsystems. What this means is that the lofty promises of a fully predictive, high-precision approach to maintenance planning is typically reserved for mega fleets – and results in noisy, low-quality forecasts for everyone else.
  • Lack of flexibility/context: Even a theoretically perfect forecast of failure probability on the component/subsystem level – for each VIN – could overlook much of the business context that goes into making solid maintenance decisions. Any good maintenance plan, for example, should take into account the warranty status of the vehicle, inventory levels, and staffing plans. Too often, predictive maintenance looks only at the specifics of the vehicle itself – without any integration of the broader business context.
  • Operational tradeoffs: It’s great to know that a crankshaft has a 43% likelihood of failure in the next 500 miles. But does that mean the driver should take the truck in for immediate inspection? Or should a tech carry out a crankshaft inspection at the next scheduled maintenance stop? This requires the notion of an expected value for each action in a given scenario, which in turn requires inputs on the specifics of your business (things like the cost of a failure or a repair). Lack of visibility into these factors – particularly on the part of, say, an OEM – makes it difficult to deliver useful “last mile” recommendations.

Relative to conventional predictive maintenance strategies, adaptive maintenance strategies tend to drive faster adoption through the company (because they’re tailored to the right level of complexity). And they produce a more sustained impact on maintenance KPIs, since they’re engineered to balance vehicle risk alongside key business requirements.

Case in point: a heavy-duty fleet drove $98 in cost savings per vehicle a year from batteries alone with an adaptive maintenance strategy by a) analyzing battery failure rates across the fleet to update maintenance intervals (for out-of-warranty vehicles) to avoid over-inspection; and b) forecasting and inspecting the highest-risk VINs for more frequent maintenance. 


Four steps to adopting an adaptive maintenance strategy

Implementing an adaptive maintenance strategy doesn’t require guesswork. The most successful fleets follow these four steps: 

  • Figure out your next best play. Assess your data and analytics maturity. Fleets tend to split on data (the type and volume of data they collect), tools (what software they use to manage business processes), and maintenance practices (general approach to optimizing maintenance plans).​​
  •  -
  • Decide where to focus. Prioritize the right components and subsystems. Primary fleet challenges tend to stem from over-maintaining (incurring unnecessary cost) or under-maintaining (increasing the likelihood and severity of incurring high costs from unplanned maintenance). But it’s not always clear where to start, which is where benchmarking data can come in handy. Services like the TMC/FleetNet benchmark enable fleets to get granular data on maintenance costs by component or subsystem to identify areas of opportunity.  
  • Don’t forget about the last mile. It’s not enough to generate machine learning predictions. You need to translate them into a language that service techs can enact in the field – typically rules, alerts, and policies. Start by mapping out the exact workflows and data flows used by team members in the field and then identifying the lowest-friction ways to integrate with them.
  • Get started…small, if you have to, but now. Inaction today makes it harder to close the gap over time. Focus on achievable steps – even if it’s just optimizing the maintenance routine for a single subsystem or component.


Closing thoughts

Predictive analytics is a great idea. But it requires adaptive maintenance – the holistic, data-driven approach to reducing maintenance costs and downtime – to move the needle on the maintenance KPIs that matter most.

Whatever approach you take, remember that time is your most valuable asset. And get started today.


About Viaduct

Viaduct's machine learning solutions allow fleets to manage, analyze, and utilize the data generated by their connected vehicles. With a platform powered by patented machine learning technology and data analyzed from over 2 million vehicles on the road, Viaduct is helping fleets overcome the biggest barriers to operationalizing machine learning at scale.

For more information, contact [email protected].